Portable 3D scanners with low measurement uncertainty are ideally suited for capturing the 3D shape of objects
right in their natural environment. However, elaborate manual post processing was usually necessary to build a
complete 3D model from several overlapping scans (multiple views), or expensive or complex additional hardware
(like trackers etc.) was needed. On the contrary, the NavOScan project aims at fully automatic multi-view
3D scan assembly through a Navigation Unit attached to the scanner.
This light weight device combines an optical tracking system with an inertial measurement unit (IMU)
for robust relative scanner position estimation. The IMU provides robustness against swift scanner movements
during view changes, while the wide angle, high dynamic range (HDR) optical tracker focused on the measurement
object and its background ensures accurate sensor position estimations. The underlying software framework,
partly implemented in hardware (FPGA) for performance reasons, fusions both data streams in real time and
estimates the navigation unit’s current pose. Using this pose to calculate the starting solution of the Iterative
Closest Point registration approach allows for automatic registration of multiple 3D scans. After finishing the
individual scans required to fully acquire the object in question, the operator is readily presented with its finalized
complete 3D model!
The paper presents an overview over the NavOScan architecture, highlights key aspects of the registration
and navigation pipeline and shows several measurement examples obtained with the Navigation Unit attached
to a hand held structured-light 3D scanner.